Traffic Light Signal Recognition Using Multi-Image Frame Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous driving systems face challenges in accurately and robustly recognizing traffic light signals, particularly due to performance degradation of sensors, bad weather, backlighting, and inefficiencies in calculation amount and speed.
Innovation Solution
A method that collects multiple images of a traffic light, extracts signal state information from each image, and determines final signal information using a pre-trained signal classification model and time-series information about the traffic light, thereby enhancing recognition accuracy and robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple images are collected and analyzed to improve recognition accuracy, then recognition accuracy and robustness are improved, but calculation amount and processing time increase
Solution Approach 1:
The patent divides the image analysis process into two stages: first, rapidly screen multiple images to identify frames containing clear traffic light signals; second, perform detailed signal state analysis only on the selected frames. This segmentation reduces the number of images requiring full analysis, thereby maintaining recognition accuracy while reducing overall calculation amount and processing time.
Solution Approach 2:
The patent applies partial action by collecting more images than traditionally required (excessive action in data collection) but then using selection criteria to process only a subset (partial action in processing). This approach ensures that sufficient data is gathered to maintain accuracy under various conditions while avoiding the computational burden of processing all collected images.
2Reliability
If the entire image is analyzed to recognize traffic light signals, then recognition completeness is improved, but calculation amount increases
Solution Approach 1:
The patent extracts and focuses analysis on specific regions of interest within the captured images, particularly areas where traffic lights are detected or expected. By applying region-of-interest (ROI) techniques, the system analyzes only the relevant portions of images rather than processing the entire image data, thereby maintaining recognition completeness while significantly reducing calculation amount and improving processing efficiency.
3Productivity
If single frame analysis is used to reduce calculation amount, then calculation efficiency is improved, but recognition accuracy and robustness deteriorate
Solution Approach 1:
The patent performs preliminary analysis on multiple images to evaluate signal quality, clarity, and detectability before committing to detailed recognition. By pre-screening images and selecting those with the most reliable traffic light signals, the system ensures that subsequent detailed analysis is performed on high-quality data, thereby maintaining recognition accuracy while avoiding unnecessary processing of poor-quality images.
Solution Approach 2:
The patent implements a feedback mechanism where the results from preliminary screening influence the selection of images for detailed analysis. Images that show clear traffic light signals based on preliminary metrics are selected for full analysis, while ambiguous or poor-quality images are either re-captured or excluded. This feedback loop ensures optimal use of computational resources while maintaining high recognition accuracy.
Data Source
AI summary
Provided are a signal information recognition method, device, and computer program for the autonomous driving of a vehicle. The signal information recognition method for the autonomous driving of a vehicle is performed by a computing device, and comprises the steps of: collecting a plurality of images generated by capturing images of a traffic light located in a prescribed area; extracting a plurality of pieces of signal state information from each of the plurality of collected images; and determining final signal information about the traffic light by using the extracted plurality of pieces of signal state information.